Marc Burri
Economics Ph.D.
[ CV ]I am an economist with a passion for data science and econometrics. I leverage alternative, unstructured, and big data for economic forecasting, using time series and machine learning techniques. My work involves macroeconomic and textual data to improve predictions and support policy-making.
Interests
Latest projects
New R Package bridgr
[ Installation ] [ Get started ] The package implements bridge models for nowcasting and forecasting macroeconomic variables by linking high-frequency indicator variables (e.g., monthly data) to low-frequency target variables (e.g., quarterly GDP). Simplifies forecasting and aggregating indicator variables to match the target frequency, enabling timely predictions ahead of official data releases.
Multi-dimensional monetary policy shocks based on heteroscedasticity
[ Working Paper ] [ Comments are very welcome! ] Joint with Daniel Kaufmann Abstract: We propose a two-step approach to estimate multi-dimensional monetary policy shocks and their causal effects requiring only daily financial market data and policy events. First, we combine a heteroscedasticity-based identification scheme with recursive zero restrictions along the term structure of interest rates to disentangle multi-dimensional monetary policy shocks and derive an instrumental variables estimator to estimate dynamic causal effects. Second, we propose to use the Kalman filter to compute the linear minimum mean-square-error prediction of the unobserved monetary policy shocks. We apply the approach to examine the causal effects of US monetary policy on the exchange rate. The heteroscedasticity-based monetary policy shocks display a relevant correlation with existing high-frequency surprises. In addition, their dynamic causal effects on the exchange rate are similar. This suggests the approach is a valid alternative if high-frequency identification schemes are not applicable.
Three Centuries of Swiss Economic Sentiment
[ Mimeo ] [ Comments are very welcome! ] Abstract: There is a lack of consistent and well-measured Swiss business cycle indicators over long historical episodes. This paper fills this gap by constructing a business cycle indicator on quarterly frequency spanning from 1820 to 2021. Using textual data such as historical company records, newspapers, and business association reports, I develop a business cycle indicator, drawing on sentiment and count-based measures related to key economic concepts. This approach involves extensive data collection, surpassing existing datasets in scope and historical coverage. The composite indicator demonstrates strong correlations with real economic activity, effectively capturing historical downturns and expansions. I also employ it to estimate recession probabilities, shedding light on Switzerland’s economic history. This paper contributes by introducing a comprehensive business cycle indicator, assembling a rich textual dataset, presenting innovative text mining methods, and establishing the first business cycle dating for Switzerland in the 19th and early 20th centuries.